0.25
2.5
probably more compute intensive if you're trying to do that with linear gradients
I guess the real question is, although blend2d is primarily focused on vector graphics, how much consideration needs to be made to deal with pixel effects in general? Is it a large departure from the vector focus, or is it a slight deviation? It seems unavoidable that more pixel processing will want to sneak its way into most applications, it starts with blt and goes from there.
And for code's sake
for (int y = 0; y < 200; y+=1) {
double gamma = map(y, 0, 200, 0.025, 1.25);
for (int x = 0; x < width - 1; x++) {
double wl = map(x, 0, width - 1, 380, 780);
auto c = coloring::ColorRGBAFromWavelength(wl, gamma);
stroke(BLRgba32(c.r * 255, c.g * 255, c.b * 255));
line(x, y, x, y+1);
}
} for (int y = 0; y < 200; y+=1) {
double gamma = map(y, 0, 200, 0.025, 1.25);
for (int x = 0; x < width - 1; x++) {
double wl = map(x, 0, width - 1, 380, 780);
auto c = coloring::ColorRGBAFromWavelength(wl, gamma);
auto p = BLRgba32(c.r * 255, c.g * 255, c.b * 255, 255);
canvas->set(x, y, p);
}
}
It might be that to get into the pipeline of multi-threaded operations, I could "reserve" a section of pixels, and perform whatever I'm going to perform, and the rendering pipeline will integrate that into the general command processing machinery. that seems like a quite reasonable thing to do. Could be used for image conversions, for example, or when I'm decoding mjpeg, or whatever.
@Wiladams yes this would be very useful. Some time ago I proposed something similar too, a mechanism to schedule an arbitrary function to be executed on a rectangular region of a BLImage, fitting it in the multithreaded flow of blend2d. Iirc @kobalicek took note of this.
And @kobalicek : arm Macs are real, and no avx support in Rosetta 2! Ouch!
@Wiladams I think we should send a container of milk and cookies to @kobalicek to fuel him for the arm porting work
@Wiladams I haven't finalized any donation system so far - yeah the web states such amount to basically avoid small individual donations by wire and to focus more on companies, and I don't know how to do the "smaller" donations properly - I have checked github sponsors and I may try that out - but that's monthly donations basically, so not sure if this is what people would prefer.
In addition, what I have thought about was creating a transparent account in EUR currency, which would allow people to donate through SEPA payments (those are pretty cheap in EU, much better than donation platforms that take 15%). I can do that with my current bank, but my experience is that people just don't want to donate, so I'm not sure it's worth the shot.
And ARM - well, that's unfortunate yeah, I have an ARM asmjit branch already, but I haven't moved that far actually
But I'm not really interested in many currencies - I think the only reasonable is EUR and USD - and USD transfers are pretty expensive
I don't know, but I know that I have stopped using PayPal years ago as it was very bad with fees - I think they charged like 5% for transfer and then another 10% for local currency conversion, they don't let me withdraw EURs, so I usually used PayPal to buy something from AliExpress instead :)
I would have to make a small research to see what would be the best way to make donations possible, the current approach is not working actually
What about this github sponsors, does it make sense? I have it almost setup, but I have no idea
I think there is even more ways of supporting the project, what about GSOC for example? Do anyone has any experience with that? Would Blend2D be a viable project for that?
Update: Still working on support for masks (fill / blit). Starting with ST implementation first, still don't have it ready. In addition, I bought the newest RPI 4 for experimenting and I can run Blend2D on ARM64, but still without JIT. However, the particles demo needs only 7ms to render 1200 particles on a screen with 1024x600 resolution, which is pretty nice considering there is no SIMD in the reference pipeline code.
Looks like that - a bit low quality image, it's already evening here :)
The important thing is, I will be starting with ARM64 support before 32-bit ARM
or start with a larger canvas, then center the result?
